Severe sepsis and septic shock

Transcription

1 Feature Articles Impact of continuous venovenous hemofiltration on organ failure during the early phase of severe sepsis: A randomized controlled trial* Didier Payen, MD, PhD; Joaquim Mateo, MD; Jean Marc Cavaillon, PhD; François Fraisse, MD; Christian Floriot, MD; Eric Vicaut, MD, PhD; for the Hemofiltration and Sepsis Group of the Collège National de Réanimation et de Médecine d Urgence des Hôpitaux extra-universitaires Objective: The impact of continuous venovenous hemofiltration on sepsis-induced multiple organ failure severity is controversial. We sought to assess the effect of early application of hemofiltration on the degree of organ dysfunction and plasma cytokine levels in patients with severe sepsis or septic shock. Design: Prospective, randomized, open, multicenter study setting, 12 French intensive care units. Patients: A total of 80 patients were enrolled within 24 hours of development of the first organ failure related to a new septic insult. Interventions: Patients were randomized to group 1 (HF), who received hemofiltration (25 ml/kg/hr) for a 96-hour period, or group 2 (C) who were managed conventionally. Measurements and Main Results: The primary end point was the number, severity, and duration of organ failures during 14 days, as evaluated by the Sepsis-Related Organ Failure Assessment score, on an intention-to-treat analysis. Strict guidelines were provided to perform continuous hemofiltration under the same conditions and bearing the same objectives in all centers. Because of inclusion stagnation, the trial was discontinued after an interim analysis by which time 76 patients had been randomized. The number and severity of organ failures were significantly higher in the HF group (p < 0.05). No modifications in plasma cytokine levels could be detected. Conclusion: These data suggest that early application of standard continuous venovenous hemofiltration is deleterious in severe sepsis and septic shock. This study does not rule out an effect of high-volume hemofiltration (>35 ml/kg/hr) on the course of sepsis. (Crit Care Med 2009; 37: ) KEY WORDS: septic shock; hemofiltration; cytokines; organ failure *See also p From the Department of Anesthesiology and Critical Care Medicine (DPJ, JM), Lariboisière Hospital, University Paris, Paris, France; Unit Cytokines & Inflammation (JMC), Institut Pasteur, Paris, France; Department of Intensive Care (FF), Delafontaine Hospital, Saint-Denis, France; Department of Intensive Care (CF), P. Morel Hospital, Vesoul, France; and Unité de Recherche Clinique (EV), Fernand-Widal University Hospital, Paris, France. Supported, in part, by the Plan Quadriennal de Ministère de la Recherche, EA 322, and from Baxter/ Edwards Life Sciences for hemofiltration materials, logistic support, and statistical analysis. Presented as an abstract at the SCCM meeting in Phoenix in Didier Payen holds a consultant contract with Edwards Life Sciences, and received a grant for research not related to the present study. The remaining authors have not disclosed any conflicts of interest. Severe sepsis and septic shock continue to pose a considerable challenge. Mortality and morbidity remain high despite increased understanding of the underlying pathophysiology and advances in technical support. Outcome relates to sepsis severity, with death mainly occurring from multiple organ failure during the first week of evolution (1). Although exact mechanisms linking infection, inflammation, and organ dysfunction are not fully understood, it is generally accepted that organ failure results as a consequence of a storm of inflammatory mediators and excessive cellular activation Authors Contributions: As principal investigators of the Hemofiltration and Sepsis Study group trial, Drs. Payen and Fraisse had full access to all data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Payen, Fraisse, Vicaut, Floriot, and Moret. Acquisition of data: Payen, Mateo, and Cavaillon. Analysis of data: Payen, Mateo, Cavaillon. Drafting of the manuscript: Payen, Mateo, and Cavaillon. Critical revision of the manuscript for important intellectual content: Payen and Cavaillon. Statistical expertise: Independent company, Vicaut, Payen, and Mateo. Study supervision: Payen. For information regarding this article, dpayen1234@aol.com Copyright 2009 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: /CCM.0b013e (2). The Sepsis-Related Organ Failure Assessment (SOFA) score (3) measured within the first 2 days of severe sepsis or septic shock is predictive of day 28 mortality (1). The recent demonstration of outcome benefit from early circulatory optimization (4) suggests that a golden window exists for early interventions that may limit organ failure progression and reduce subsequent mortality. By contrast, late optimization, performed after organ failure is now established, has no beneficial effects, and may even be deleterious (5). Because of its simplicity and efficiency in taking over for the failing kidney, continuous venovenous hemofiltration (CVVH) is widely used to manage renal failure (6). However, CVVH is frequently delayed for several days until sufficiently abnormal plasma variables develop to motivate the initiation of renal replacement therapy. In addition to the classic indications for renal support, such as azotemia, fluid overload, acidosis, and hyperkalemia, some experimental and clinical studies suggest a potential benefit through mod- Crit Care Med 2009 Vol. 37, No

2 ifying the plasma patterns of inflammation (7 12) and organ failure (13). This concept gained further credibility after the publication by Honore et al (14) suggesting benefit from short-term highvolume hemofiltration in patients with refractory septic shock, particularly when applied early. The current study reports the results of a multicenter trial assessing the classic hemofiltration technique (i.e., not utilizing high-volume fluid removal) applied early in severe sepsis or septic shock, and its effects on the number and severity of organ failures. METHODS Study Design and Organization. The Ethical Committee of Health Care of Poitou- Charentes approved this randomized, controlled, multicentered trial (number ), which was regulated by an Independent Data Safety and Monitoring Board. Patients were enrolled between January 1997 and January 2000 in the intensive care units of two university and ten community hospitals. Informed consent was obtained from the next of kin of each patient. None of the centers received financial support from industry. Edwards Life Sciences (Irvine, CA) provided hemofilters for use in the study and an unrestricted grant to allow study monitoring and statistical analyses by an independent company. Patients. The following inclusion criteria were used according to the American College of Chest Physicians/Society of Critical Care Medicine consensus conference (15): clinically identified focus on infection associated with at least 2 systemic inflammatory response syndrome criteria and one or more sepsis-induced organ failures within the 24 hours before inclusion, plus a Simplified Acute Physiology II score between 35 and 63. Patients were excluded if pregnant, younger than 18 years, in a moribund state, in chronic renal failure, or receiving immunosuppressive therapy. Randomization. Patients meeting the criteria were randomized by blocks of four to receive 96 hours of isovolemic CVVH in addition to standard sepsis management in one group, or standard sepsis management alone in the other group. The protocol had to start within 24 hours after randomization, thus allowing sufficient time to confirm the presence of infection and to optimize resuscitation. Data were collected daily during the first 5 days, then at days 7, 10, and 14. The primary end point was the occurrence or the worsening of sepsis-induced organ failures as evaluated by the SOFA score measured from day 0 to day 14. Day 0 (the day of inclusion into the study) was used as the reference value for subsequent evolution of the SOFA score and scored as 0. An event was defined by a variation of 1 point for at least two organ systems. Secondary end points comprised mortality rate at day 14 and time to death. Time to withdrawal of catecholamines and time to weaning from mechanical ventilation were also analyzed. Because the SOFA score reduction has not been evaluated from known data to predict outcome, the sizing of the population was difficult to determine a priori. Practically, the study was powered to fulfill the secondary end point of mortality rate. The sample size was fixed at n 400 patients to ensure an 80% power to detect a 20% decrease in mortality, considering a 48% mortality rate in one arm of the study group. Hemofiltration Technique. The same CVVH technique was used in all participating centers. Vascular access was via a 14F duallumen venous catheter (Arrow, Reading, PA) inserted mainly in the femoral vein. A heparin-coated polysulfone membrane (Duraflo II, Edwards Life Sciences, Irvine, CA) of 1.2 m 2 surface area was used. The cutoff coefficient of this membrane is 30,000 Da. The circuit blood flow was set at 150 ml/min. The ultrafiltration rate was set at a fixed 2 L/hr for all patients. Warmed replacement fluid was bicarbonate buffered (32 mmol/l) (Hemosol B0, Hospal, Lyon, France) delivered at a rate of 2 L/hr. Net fluid removal was set by the physician in charge according to the patient s condition and clinical need. Total patient fluid balance was calculated for the 14-day study period (Table 4). In the CVVH group, the duration of hemofiltration was at least 96 hours. To ensure the use of efficient membranes, the hemofilter was replaced at 12, 24, and 48 hours, and at other recorded times if the filter clotted prematurely. Blood circuit anticoagulation consisted of prefilter administration of heparin sodium (Heparine; Choay; Paris, France) at a rate of 5 IU/kg/hr. Glucose and potassium chloride was optionally added to the 10-L bag of replacement fluid according to patient s need. Baseline Assessment and Data Collection. The Simplified Acute Physiology II score was calculated within the first 24 hours after meeting the inclusion criteria. SOFA was calculated at the time of enrollment using the most aberrant values, and daily thereafter for the next 14 days until death or discharge. The centers were free to use the pulmonary artery catheter inserted as per protocol for data collection to enable additional cardiovascular monitoring and for choice and titration of drug therapy. The variables required to define organ failure, including routine laboratory constants, were collected daily at 8 AM. The decision to modify cardiovascular support and to wean from mechanical ventilation was made by the clinical team caring for the patient. The duration of these supportive therapies was accurately recorded (days and hours), referring to the precise time of protocol inclusion. The length of intensive care unit stay (days) was calculated using the day at which a general ward bed was first requested. Cytokine Measurements. Two centers participated in this substudy. Blood samples (5 ml) were drawn into ethylenediaminetetraacetic acid, centrifuged, and 1 ml plasma aliquots were kept at 80 C until assayed for cytokine content. Samples were obtained upstream and downstream of the hemofilter at 0, 1, and 12 hours after initiation of hemofiltration, 1 hour after replacement of the first hemofilter (13 hours), and 24 hours. Samples of ultrafiltrate were collected at the same time points. For the nonhemofiltered patients, samples were collected at inclusion, and 12 and 24 hours later. Eleven markers were measured, i.e., tumor necrosis factor alpha, interleukin (IL)-6, IL-8 (or CXCL8), IL-10, IL-1 receptor antagonist (IL-1ra), oncostatin M, monocyte chemoattractant protein-1 (MCP-1 or CCL2), macrophage inflammatory protein (MIP-1 or CCL4), regulated upon activation normal T-cell expressed and secreted (RAN- TES or CCL5), soluble tumor necrosis factor receptor II, and the soluble IL-6 receptor. Specific enzyme-linked immunosorbent assays from R&D Systems (Abingdon, UK) were employed according to the manufacturer s instructions. Adverse events (E) were recorded throughout the study, coded using prespecified categories and then summarized according to severity and resolution. The number and percentage of patients experiencing at least one adverse event were analyzed. In any given category, subjects were only counted once. If the same event for a patient had two different severities, the worst was taken into account. Statistical Analysis. The population was analyzed on an intention-to-treat basis. All randomly assigned subjects in the hemofiltration (HF) group who received at least 1 day of hemofiltration and all patients from the control group were analyzed. Continuous variables were compared between treatments using a two-way analysis of variance. Categorical data were compared using a Fisher s exact test. Efficacy analyses were performed twice to check the sensitivity of the results to dropouts and unavailable assessments. In the analysis of the primary end point, missing data were replaced by carrying forward the last on-therapy value available before discontinuation (last observation carried forward approach). Results were compared with those obtained from analysis of all observed data. The time to multiple organ failure evolution of the two groups SOFA scores was compared by a log-rank test and were represented by a Kaplan-Meier curve. A similar method was used for analysis of time to worsening for a single organ. The death rate over time was compared within the two groups using the log-rank test and also represented by Kaplan-Meier curves. Statistical analyses were performed using SAS software (Cary, NC). Statistical testing was two-sided and used the 5% significance level in accordance with standard practice. 804 Crit Care Med 2009 Vol. 37, No. 3

3 RESULTS 37 Assigned to 96 hours isovolemic CVVH (HF Group) 0 Lost to Follow-up 80 Patients Assessment for Eligibility 76 Randomized 4 excluded 2 did not have sepsis 2 family decisions to withdrawfromthe study despite their initial informed consent 39 Assigned to Standard management of severe sepsis (Control Group) 0 Lost to Follow-up 37 Included in Analysis 39 Included in Analysis Figure 1. Patient flow diagram. CVVH, continuous veno-venous hemofiltration; HF, hemofiltration. Crit Care Med 2009 Vol. 37, No. 3 Characteristics of the Patients. From January 1997 to June 1999, 80 patients were enrolled in 12 intensive care units. Because inclusion rate showed stagnation in inclusions, the Comite d Analyse des Donnees Cliniques decided to perform an intermediate analysis. In view of the results, the committee decided to stop the trial. The Data Safety Monitoring Board Committee carefully reviewed each case record form and decided to remove four patients because of major violations of the protocol: two of them did not have a sepsisrelated inflammation and two for whom the protocol was stopped within 48 hours of inclusion because of a family decision to suspend therapy, despite their initial informed consent. As a consequence, 76 patients were analyzed (Fig. 1). Thirty-seven patients in the HF group and 39 patients in the control group fulfilled the inclusion criteria. The two groups were comparable for age, gender, sepsis etiology, baseline Simplified Acute Physiology Score, and SOFA score (Table 1), and did not differ at baseline for mean blood pressure, heart rate, PAO 2 /FIO 2 ratio, urinary output and creatinine, platelets, and plasma bilirubin (Table 2). Table 3 shows the sepsis characteristics and therapeutic interventions. The source of sepsis, the time to inclusion from the diagnosis of sepsis, the site of infection, and the bacteriologic data were comparable in the two groups. Timing for septic events was also comparable for the two groups (Table 3). The delay from systemic inflammatory response syndrome and the inclusion was hours and hours (hemofiltration group vs. conventional group). The first organ failure related to the sepsis appears (hemofiltration) and hours (conventional group) before inclusion, a delay considered early for severe sepsis. Therapeutic interventions did not differ in the two groups except for continuous renal replacement therapy (Table 3). Primary End Point. Figure 2 shows the Kaplan-Meier curves for the time to worsening of the SOFA score, as defined in the Methods section. The HF group elicited a more rapid deterioration than the control group (log-rank test: chisquare 8.73; p ). This difference remained significant until day 10. When compared to the day 0 reference value (0), the SOFA score evolution profile was statistically different in the two groups with the score in the HF group remaining higher (p 0.001). Secondary Outcomes. The secondary end point analyses showed that mortality at day 28 was not significantly different in the two groups (54% HF vs. 44% controls; p 0.49). Time to death tended, but not significantly, to be shorter in the HF group over the 14-day study period (log-rank test: chisquare 2.638; p 0.104) (Fig. 3). Figure 4 Kaplan-Meier curves showed that weaning from both mechanical ventilation (Fig. 4A) and catecholamines (Fig. 4B) was significantly longer in the HF group. For both, weaning was significantly longer in the HF group (log-rank test: chi-square 4.19, p for mechanical ventilation; chi-square 3.91, p for catecholamine use). Urinary output within the 24 hours preceding inclusion was ml in HF group; in conventional group (not significant). For assessment of renal failure, we used only a single item of the SOFA definition, which normally considers abnormalities of either plasma creatinine or urinary output. We chose a urinary output lower than 500 ml/day, because the HF group was supposed to have a maintained or reduced creatinine level. Using this urinary output criterion, the frequency of renal failure was higher in the HF group (analysis of variance; p 0.05), although the proportion of patients in renal failure was similar at day 0 (Table 2) in both groups. As a consequence, the use of HF for renal support after the 96 hours of protocolized HF was significantly higher in the HF group (19 of 37 in the HF group vs. 8 of 39 in the control group; p 0.05). Evolution over 14 days of the relevant score values for organ failure showed that hemodynamic variables, PAO 2 /FIO 2 ratio, plasma creatinine, and fluid balance did not differ in the two groups and fell within an adequate range for resuscitation goals. 805

4 Table 1. Baseline demographic characteristics of the study patients Characteristics HF Group (n 37) Control Group (n 39) p Age (yrs) NS Female sex, n (%) 10 (27) 12 (31) NS Simplified Acute Physiology II score NS Body weight (kg) NS Sequential Organ Failure Assessment score NS Type of patient Scheduled surgery, n (%) 1 (3) 2 (5) Medical, n (%) 19 (51) 25 (64) Emergency surgical, n (%) 17 (46) 12 (31) Infection status Systemic inflammatory response syndrome duration before inclusion (hrs) NS NS Organ failure duration before inclusion (hrs) NS Infection sites NS Catheter, n (%) 0 (0) 5 (10) Abdominal, n (%) 14 (33) 11 (23) Respiratory, n (%) 14 (33) 18 (38) Urinary, n (%) 6 (14) 2 (4) Other, n (%) 8 (19) 12 (25) HF, hemofiltration; NS, not significant. Table 2. Baseline clinical characteristics of the study patients Characteristics HF Group (n 37) Control Group (n 39) p Hemodynamic Heart rate (beats/min) NS Systolic arterial pressure (mm Hg) NS Diastolic arterial pressure (mm Hg) NS Mean arterial pressure (mm Hg) NS Systolic pulmonary arterial pressure (mm Hg) NS Diastolic pulmonary arterial pressure (mm Hg) NS Mean pulmonary arterial pressure (mm Hg) NS Pulmonary artery occlusion pressure (mm Hg) NS Right atrial pressure (mm Hg) NS Cardiac output (L/min) NS Respiratory PaO 2 /FIO NS PaCO 2 (mm Hg) NS Positive end-expiratory pressure (cm of water) ph NS Lactate (mmol/l) NS Platelets (10 9 /L) NS Prerandomization fluid input (ml) NS within 24 hrs Renal Urinary outputs within the 24 hrs NS preceding inclusion (ml) Prerandomization diuresis within NS 24 hrs (L/24 h), n (%) 1 L 19 (27.1) 19 (27.1) L 9 (12.8) 10 (14.2) 0.5 L 6 (8.5) 7 (10) Urea (mmol/l) NS Creatinine ( mol/l) NS Others Glasgow Coma score NS Bilirubin, ( mol/l) NS Prothrombin level (%) NS HF, hemofiltration; NS, not significant. The actual dose used over 96 hours was in total ml/hr or 24.7 ml/kg/hr (Table 4). The timing of hemofilter membrane replacement and the desired ultrafiltration rate were achieved for all patients. Within the 96 hours of the HF protocol, fluid removal averaged 120 ml/hr (range ml/hr). Net fluid balance was not statistically different during the study period (81 78 kg in the HF group, kg in controls) (Table 3). There were no differences in adverse events within the two groups. Cytokines. Plasma cytokine measurements were achieved in ten HF and eight control patients, and demonstrated the anticipated elevated levels as classically seen in septic shock. There were no statistical differences for all the cytokines measured between the two groups before inclusion. For all markers, the levels upstream and downstream of the hemofilter were similar except for IL-6, MCP-1, IL1- ra, and IL-10, which increased slightly in relation to hemoconcentration (data not shown). Sequential measurements at times T0, T12, and T24 hours did not display any significant differences between groups (Fig. 5). Some markers (IL- 1ra, IL-6, IL-8, and MCP-1) could be detected in the ultrafiltrate at concentrations up to 35% of the plasma levels ones. There was a significant correlation between plasma and ultrafiltrate levels (Fig. 6). In contrast, zero or very low levels ( 1% of plasma levels) of RANTES, MIP-1, IL-10, soluble IL-6 receptor, and soluble tumor necrosis factor receptor II could be measured in the ultrafiltrate. After grouping patients on outcome and not on studied groups, we observed differences for procalcitonin, IL-10, IL-6, and RANTES, which were significantly higher in dead patients than in survivals at inclusion time. DISCUSSION Early application of classic CVVH to a population with severe sepsis, irrespective of the presence of renal failure, failed to limit or improve organ failure. Contrary to expectation, it prolonged the requirement for organ support and also showed a trend to higher mortality at 14 days compared with a control group, requiring discontinuation of the study as justified by fulfillment of a priori determined stopping rules. No modifications in cytokine plasma levels could be detected. The two groups were well balanced in terms of demographics, bacteriology, physiology, and illness severity. Although impossible to perform a double-blinded study using this device, importantly, the potential bias induced by the nonblinded protocol seems limited because cardiorespiratory and metabolic data did not differ between the two groups over the study 806 Crit Care Med 2009 Vol. 37, No. 3

5 Table 3. Intervention and support intensity Figure 2. Kaplan-Meier curves for time to worsening of the Sepsis-Related Organ Failure Assessment (SOFA) score. The hemofiltration (HF) group showed a more rapid deterioration compared with the control (C) group (log-rank test: ; p 0.01). Figure 3. Kaplan-Meier curves for time to death during the observation period (14 days). Logrank test: ; p C, control; HF, hemofiltration. period. Furthermore, the goals for resuscitation achieved in both groups were appropriate, albeit with a greater intensity when using supportive techniques in the CVVH group. The dose-ranging study HF Group (n 37) Control Group (n 39) p Arterial catheter, n (%) 31 (83.7) 30 (76.9) NS Central catheter, n (%) 34 (94.4) 36 (92.3) NS Pulmonary artery catheter, n (%) 18 (48.6) 16 (41) NS Mechanical ventilation, n (%) 35 (94.5) 36 (92.3) NS Vasopressors, n (%) Dopamine, n (%) 9 (25) 14 (36) NS Dobutamine, n (%) 9 (25) 10 (26) NS Epinephrine, n (%) 5 (14) 6 (15) NS Norepinephrine, n (%) 24 (39) 21 (54) NS Steroids 1 (3) 2 (5) NS Antibiotics 100% 100% NS HF, hemofiltration; NS, not significant. Crit Care Med 2009 Vol. 37, No. 3 Figure 4. A, Kaplan-Meier curves for time to weaning from mechanical ventilation. Weaning took a significantly longer time in the hemofiltration (HF) group (log-rank test: ; p 0.04). B, Kaplan-Meier curves for time to weaning of catecholamines. Time to weaning was significantly longer in the HF group (log-rank test 2 3.9; p 0.048). C, control. from Ronco et al (13), suggestive of improved outcomes when using ultrafiltrate rates of 35 ml/kg/hr, had not been published before the design of our protocol which targeted a flow rate of approximately 25 ml/kg/hr. Because our trial did not focus on renal support, dose ranging might have had a more limited impact on the measured variables. However, Cole et al (16) clearly demonstrated that among patients with septic shock, the dose of norepinephrine needed to reach the target blood pressure was significantly less in patients treated with high-volume hemofiltration (6 L/hr) than during conventional hemofiltration (1 L/hr). However, this study was not designed to draw conclusions about outcome. Regular filter membrane changing might have limited any time-related performance deterioration during the 96- hour study period. Finally, any heterogeneity in technical performance between intensive care units was negligible, as no severe adverse events related to CVVH were observed and this was applied early (mean 22 hours) during the course of severe sepsis. The clinical use of hemofiltration to limit inflammation, already suggested by experimental data (17), gained even more credibility after the study from Honore et al (14) in which high-volume hemofiltration rescued 45% of patients in refractory septic shock, despite a catastrophic predicted mortality. The technique required to achieve such a goal may differ from the standard procedure in relation to hemofiltration volume, the use of high permeability membranes, or special circuits and cartridges to adsorb molecules of inflammation (10, 16, 18 23). Several small trials have shown the superiority of highpermeability membranes in terms of vasopressor reduction (24), restoration of mononuclear cell proliferation, and attenuation of polymorphonuclear neutrophil phagocytosis (7, 25 28). For the primary end point, we used the well-established SOFA score (3) to assess any potential benefit of CVVH on organ failure. The day-to-day use of this score for such a purpose has recently been validated as a prognostic tool in severe sepsis (1). Application of CVVH delayed the organ failure improvement that we observed in the control group (p 0.001) (figure not shown). The time concordance between the absence in organ failure improvement and application of hemofiltration suggests a direct relationship between CVVH and prolongation of organ failure. The absence of any benefit of classic hemofiltration to limit organ failure and remove mediators has been suggested by a small phase II trial in patients with severe sepsis or septic shock (29). They did not observe any benefit at 48 hours of hemofiltration using a 2-L hourly exchange. Our two groups have 807

6 Table 4. Hemofiltration protocol used; average value and min/max range Factor Average Value Minimum/Maximum Blood flow rate (ml/min) 159 ( ) Dose of hemofiltration (ml/hr) ( ) Over 96 hrs Indexed dose (ml/kg/hr) 24.7 ml/kg/hour Hourly fluid removal (ml) 119 (48 322) Total fluid removal (ml/24 hr) 2715 ( ) Hemofilter circuit clotting 96 hrs 20.5% 14 days 16.81% Clotting of venous return circuit 4.5% Figure 5. Circulating cytokine levels in hemofiltration and contol groups at T0, T12, and T24 hours poststudy enrollment. IL, interleukin; MCP, monocyte chemoattractant protein. Figure 6. Correlation between plasma and ultrafiltrate concentrations of interleukin (IL)-6, monocyte chemoattractant protein (MCP)-1, and IL-1ra. similar severity and organ failure patterns, eliminating any bias inpatient recruitment to the HF group. Compared with the study by Cole et al (29), the current study not only failed to show any positive impact of hemofiltration, but rather showed a deterioration. Although no clear explanations support our observation, some suggestions can be made. Hemodynamic, respiratory, and biological values in both groups met recommended goals for resuscitation of patients with severe sepsis (30). This attenuated the risk of any bias in resuscitation of one group compared with the other. Nevertheless, the type of filter membrane used may have had an impact. For example, polyacrylonitrile (AN69) has a proven high hemoadsorption capacity (31), whereas polysulfone filters do not perform as well in this regard, even though the membranes were frequently replaced. Comparison of secondary end points also revealed differences between the two groups. The requirement for supportive therapy was significantly higher in the HF group with a longer duration of mechanical ventilation, cardiovascular and renal supports. Furthermore, a trend to an increased mortality was observed in the HF group. Because the current study was designed to not use HF as renal replacement therapy, the occurrence of renal failure should also be considered as part of its impact on organ failure. Because the RIFLE score had not been implemented when the protocol was designed, this score was not used for renal failure classification. In addition, the context of such a study, i.e., the use of HF for no renal effectiveness, renal failure scores were difficult to apply. Because HF modifies renal failure parameters, such as serum urea and creatinine, only urinary output was considered to characterize renal failure. The frequency of low urinary output was higher in the HF group. More importantly, the need for renal failure support after the 96-hour protocol had elapsed was also statistically greater in the HF group. Thus, HF seems to have a negative impact even in renal dysfunction. It is conceivable that HF may have a negative impact on renal failure, leading to the hypothesis that HF may worsen the condition of the septic stressed kidney. Such an effect does not seem related to clear hemodynamic or fluid balance differences, as both groups were comparable at the time of study enrollment. Early HF may potentially remove nonidentified 808 Crit Care Med 2009 Vol. 37, No. 3

7 small molecules that have a protective role. Furthermore, the need for more intense cardiovascular support may also worsen the kidney s functional capability. Although cytokine plasma levels do not reflect tissue cytokine content, a lower value could be expected during HF. In the current study, HF did not show any impact on plasma cytokine levels. The patterns observed were abnormal and were consistent with those previously observed in severe sepsis or septic shock (32). Previous studies have shown an early reduction in cytokine levels after commencement of HF followed, within hours, by an increase to baseline levels (7, 33). This is thought to be related to initial adsorption of cytokines onto the membrane, which then becomes saturated. To our knowledge, no other studies have assessed the effect of HF on such a large panel of cytokines during sepsis. Levels measured upstream and downstream from the site of hemofiltration were similar, and independent of the time of sampling within the first 24 hours (data not shown). Kinetics of plasma cytokines and soluble cytokine receptors were similar whether septic patients received HF or not. Surprisingly, not all markers were found in the ultrafiltrate, and this was independent of their molecular weights. We had deliberately included assays for four chemokines chosen because of their low molecular weights, of which, some were rarely detected in ultrafiltrate despite their presence in large amounts in plasma (e.g., RANTES), whereas others could be found in all samples (e.g., MCP-1). The observed correlation between plasma and ultrafiltrate levels for some cytokines might infer that higher rates of hemofiltration could have an impact on such inflammatory mediators. Highvolume hemofiltration may be more efficient in severe sepsis, as previously suggested (14, 34). The later study confirmed the efficiency of high-volume hemofiltration on cytokine removal in comparison with conventional hemofiltration, but with no information on outcome and organ failure. Such a concept remains to be proven by a multicenter randomized trial. CONCLUSION In summary, we report the results of a multicenter trial on the use of early hemofiltration in severe sepsis aiming at limitation of organ failure. Contrary to our expectations, early HF resulted in worse outcomes and prolongation of organ support. Although hemofiltration has not proved its therapeutic efficacy in sepsis, it remains an extrarenal purification technique of choice in acute renal failure, particularly indicated in hemodynamically unstable intensive care patients. New techniques in the area of extrarenal purification are currently being developed, based on the regeneration of high ultrafiltrate volume after passing over resins that absorb excess inflammatory mediators. If practical application of the technique can be achieved, and if it remains financially viable, very high-flux techniques may eventually exert a beneficial effect on the evolution of severe sepsis and septic shock. To conclude, in septic patients, hemofiltration with an ultrafiltration rate of 2 L/hr did not limit organ failure. Whether high-volume hemofiltration is able to limit organ failure in these patients remains to be investigated. ACKNOWLEDGMENTS We thank J. Kellum and M. Singer for their advice during the preparation of the manuscript. We thank Patrick Lambert from Baxter/Edwards Life Sciences for his help for data and statistical analysis during this work. We also thank Catherine Fitting and Dr. Minou Adib-Conquy (Unit Cytokines & Inflammation, Institut Pasteur) for valuable contribution to cytokine analysis. The following members constitute the Hemofiltration and Sepsis Trial Independent Monitoring Committee Data Safety and Monitoring Board: Jean Claude Chevrolet, MD, Hôpital Universitaire de Genève; Claude Perret, MD, Centre Hospitalier Universitaire Vaudois, Lausanne; Jean Carlet, MD, Hôpital Saint-Joseph, Paris. Scientific Committee: Didier Payen, Christian Floriot, Eric Vicaut, and Georges Moret. Clinical Investigators and Sites: Roland Berard, Paul Desbief Hospital, Marseille; Paul Bouletreau, Edouard Herriot Hospital, Lyon; Frederic Delille, Centre Hospital, Montargis; Didier Dorez, Centre Hospital, Annecy; Christian Floriot, De La Haute-Saone Hospital, Vesoul; François Fraisse, Delafontaine Hospital, St Denis; Claire Gatecel, General Hospital, Beziers; Jean-Louis Pallot, Intercommunal Hospital, Montreuil; Didier Payen, Lariboisière Hospital, Paris; Monique Plaisant, Marseille 1 Hospital, Marseille; Jean-Louis Ricome, Centre Hospital, St Germain En Laye; Gerard Trouillet, Centre Hospital, Pontoise. REFERENCES 1. Levy MM, Macias WL, Vincent JL, et al: Early changes in organ function predict eventual survival in severe sepsis. Crit Care Med 2005; 33: Annane D, Bellissant E, Cavaillon JM, et al: Septic shock cytokine cascade in sepsis. Lancet 2005; 365: Vincent JL, Moreno R, Takala J, et al: The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/ failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine. Intensive Care Med 1996; 22: Rivers E, Nguyen B, Havstad S, et al: Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med 2001; 345: Gattinoni L, Brazzi L, Pelosi P, et al: A trial of goal-oriented hemodynamic therapy in critically ill patients. SvO 2 Collaborative Group. N Engl J Med 1995; 333: Ronco C, Bellomo R: Basic mechanisms and definitions for continuous renal replacement therapies. Int J Artif Organs 1996; 19: Bellomo R, Tipping P, Boyce N: Tumor necrosis factor clearances during veno-venous hemodiafiltration in the critically ill. ASAIO Trans 1991; 37:M322 M De Vriese AS, Vanholder RC, Pascual M, et al: Can inflammatory cytokines be removed efficiently by continuous renal replacement therapies? Intensive Care Med 1999; 25: Grooteman MP, Groeneveld AB: A role for plasma removal during sepsis? Intensive Care Med 2000; 26: Kellum JA, Song M, Venkataraman R: Hemoadsorption removes tumor necrosis factor, interleukin-6, and interleukin-10, reduces nuclear factor-kappab DNA binding, and improves short-term survival in lethal endotoxemia. Crit Care Med 2004; 32: Ponikvar R: Blood purification in the intensive care unit. Nephrol Dial Transplant 2003; 18:v63 v Tetta C, Bellomo R, Inguaggiato P, et al: Endotoxin and cytokine removal in sepsis. Ther Apher 2002; 6: Ronco C, Bellomo R, Homel P, et al: Effects of different doses in continuous veno-venous haemofiltration on outcomes of acute renal failure: A prospective randomised trial. Lancet 2000; 356: Honore PM, Jamez J, Wauthier M, et al: Prospective evaluation of short-term, highvolume isovolemic hemofiltration on the hemodynamic course and outcome in patients with intractable circulatory failure resulting from septic shock. Crit Care Med 2000; 28: Crit Care Med 2009 Vol. 37, No

8 15. Levy MM, Fink MP, Marshall JC, et al: 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med 2003; 31: Cole L, Bellomo R, Journois D, et al: Highvolume haemofiltration in human septic shock. Intensive Care Med 2001; 27: Tetta C, Gianotti L, Cavaillon JM, et al: Coupled plasma filtration-adsorption in a rabbit model of endotoxic shock. Crit Care Med 2000; 28: Berlot G, Di Capua G, Nosella P, et al: Plasmapheresis in sepsis. Contrib Nephrol 2004; 144: Honore PM, Joannes-Boyau O: High volume hemofiltration (HVHF) in sepsis: A comprehensive review of rationale, clinical applicability, potential indications and recommendations for future research. Int J Artif Organs 2004; 27: Nalesso F: Plasma filtration adsorption dialysis (PFAD): A new technology for blood purification. Int J Artif Organs 2005; 28: Rogiers P: High-volume hemofiltration in septic shock. Crit Care 2005; 9: Ronco C, Brendolan A, d Intini V, et al: Coupled plasma filtration adsorption: Rationale, technical development and early clinical experience. Blood Purif 2003; 21: Yonekawa M, Rogiers P, Nalesso F, et al: Cytokine and endotoxin removal in critically ill patients. High-volume hemofiltration in septic shock; plasma filtration adsorption dialysis (PFAD): A new technology for blood purification; high volume hemofiltration (HVHF) in sepsis: A comprehensive review of rationale, clinical applicability, potential indications and recommendations for future research. Coupled plasma filtration adsorption: Rationale, technical development and early clinical experience. Ther Apher Dial 2005; 9:A Morgera S, Haase M, Kuss T, et al: Pilot study on the effects of high cutoff hemofiltration on the need for norepinephrine in septic patients with acute renal failure. Crit Care Med 2006; 34: DiScipio AW, Burchard KW: Continuous arteriovenous hemofiltration attenuates polymorphonuclear leukocyte phagocytosis in porcine intra-abdominal sepsis. Am J Surg 1997; 173: Morgera S, Rocktaschel J, Haase M, et al: Intermittent high permeability hemofiltration in septic patients with acute renal failure. Intensive Care Med 2003; 29: Hoffmann JN, Hartl WH, Deppisch R, et al: Hemofiltration in human sepsis: Evidence for elimination of immunomodulatory substances. Kidney Int 1995; 48: Tetta C, Cavaillon JM, Schulze M, et al: Removal of cytokines and activated complement components in an experimental model of continuous plasma filtration coupled with sorbent adsorption. Nephrol Dial Transplant 1998; 13: Cole L, Bellomo R, Hart G, et al: A phase II randomized, controlled trial of continuous hemofiltration in sepsis. Crit Care Med 2002; 30: Dellinger RP, Carlet JM, Masur H, et al: Surviving sepsis campaign guidelines for management of severe sepsis and septic shock. Intensive Care Med 2004; 30: Rogiers P, Zhang H, Pauwels D, et al: Comparison of polyacrylonitrile (AN69) and polysulphone membrane during hemofiltration in canine endotoxic shock. Crit Care Med 2003; 31: Cavaillon JM, Adib-Conquy M, Fitting C, et al: Cytokine cascade in sepsis. Scand J Infect Dis 2003; 35: Bellomo R, Tipping P, Boyce N: Continuous veno-venous hemofiltration with dialysis removes cytokines from the circulation of septic patients. Crit Care Med 1993; 21: Ghani RA, Zainudin S, Ctkong N, et al: Serum IL-6 and IL-1-ra with sequential organ failure assessment scores in septic patients receiving high-volume haemofiltration and continuous venovenous haemofiltration. Nephrology (Carlton) 2006; 11: Crit Care Med 2009 Vol. 37, No. 3